The long-term objective of this project is to develop powerful and computationally-efficient statistical methods for statistical modeling of high-dimensional genomic data motivated by important biological problems and experiments. The specific aims of the current project include developing novel survival analysis methods to model the heterogeneity in both patients and biomarkers in genomic studies and developing robust survival analysis methods to analyze high-dimensional genomic data. The proposed methods hinge on a novel integration of methods in high-dimensional data analysis, theory in statistical learning and methods in human genomics. The project will also investigate the robustness, power and efficiencies of these methods and compare them with existing methods. Results from applying the methods to studies of ovarian cancer, lung cancer, brain cancer will help ensure that maximal information is obtained from the high-throughput experiments conducted by our collaborators as well as data that are publicly available. Software will be made available through Bioconductor to ensure that the scientific community benefits from the methods developed.